State-of-the-Art Instance Matching Methods for
              Knowledge Graphs?

Alex Boyko1,2[0000−0003−3592−4986] , Siamak Farshidi2[0000−0001−6139−921X] , and
                     Zhiming Zhao2[0000−0002−6717−9418]
                   1
                      Vrije Universiteit Amsterdam, The Netherlands
                               o.y.boyko@student.vu.nl
                    2
                      Universiteit van Amsterdam, The Netherlands
                             {s.farshidi, z.zhao}@uva.nl


        Abstract. Instance matching has attracted a wide range of research
        attentions. A systematic literature review captures knowledge regarding
        the state-of-the-art systematically, to analyze and report it in the form of
        reusable knowledge. It is difficult to compare the performance of different
        instance matching methods, even when the same benchmarking dataset
        is used.


1     Introduction
Instance matching identifies instances from different data sources that refer to
the same real-world entity [7]. It is difficult to identify the state-of-the-art solu-
tion since the every instance matching approach is tailored for specific data and
its properties.

2     Overview of the State of the Art Approaches
Existing knowledge graph instance matching approaches rely on embeddings to
represent data in the form of vectors. They aim to capture the semantic meaning
of the data by placing semantically similar inputs close together in the vector
space. Figure 1 shows the output of a literacture study on instance matching
approaches.

3     Conclusion
This study reviews the latest instance matching research. The approaches are
often tested on different benchmarking datasets. Importantly, even when the
same dataset is used, there is no single instance matching approach that performs
well with every metric. Since some methods perform well on one subset and worse
on the others, it is difficult to compare their performance. Future research can
possibly explore ways for reducing human feedback while preserving its benefits,
for example by combining supervised and self-supervised approaches.
?
    Copyright © 2021 for this paper by its authors. Use permitted under Creative
    Commons License Attribution 4.0 International (CC BY 4.0).
2       A. Boyko et al.




Fig. 1. Generalization of the workflow pipelines: RNM [7], CEAFF [6], DGMC [1],
COTSAE [4], BERT-INT [3], SelfKG [2], UEA [5].


Acknowledgement
This work has been partially funded by the European Union’s Horizon 2020
research and innovation programme, by the project of ARTICONF (825134),
ENVRI-FAIR (824068) and BLUECLOUD (862409).


References
1. Fey, M., Lenssen, J.E., Morris, C., Masci, J., Kriege, N.M.: Deep graph matching
   consensus. ArXiv abs/2001.09621 (2020)
2. Liu, X., Hong, H., Wang, X., Chen, Z., Kharlamov, E., Dong, Y., Tang, J.: A self-
   supervised method for entity alignment (2021)
3. Tang, X., Zhang, J., Chen, B., Yang, Y., Chen, H., Li, C.: Bert-int: A bert-based
   interaction model for knowledge graph alignment. In: IJCAI. pp. 3174–3180 (2020)
4. Yang, K., Liu, S., Zhao, J., Wang, Y., Xie, B.: Cotsae: Co-training of structure and
   attribute embeddings for entity alignment. In: AAAI (2020)
5. Zeng, W., Zhao, X., Tang, J., Li, X., Luo, M., Zheng, Q.: Towards entity alignment
   in the open world: An unsupervised approach. In: DASFAA (2021)
6. Zeng, W., Zhao, X., Tang, J., Lin, X., Groth, P.T.: Reinforcement learning based
   collective entity alignment with adaptive features. ArXiv abs/2101.01353 (2021)
7. Zhu, Y., Liu, H., Wu, Z., Du, Y.: Relation-aware neighborhood matching model for
   entity alignment. In: AAAI (2021)